Load Data

analysis_data = readRDS(here("cleaned_data.rds")) %>% mutate_if(is.character, factor)
var_cats = read_csv(here("names_to_keep.csv"))
## Parsed with column specification:
## cols(
##   name = col_double(),
##   value = col_character(),
##   keep = col_logical(),
##   discrim = col_logical(),
##   outcome = col_logical()
## )
demo_names = filter(var_cats, discrim)$value %>% .[. %in% names(analysis_data)]
outcome_names = filter(var_cats, outcome)$value %>% .[. %in% names(analysis_data)]

no_demo_data = analysis_data %>% select(-all_of(demo_names))
outcome_data = analysis_data %>% select(all_of(outcome_names))

Descriptive Statistics

descr_stats = analysis_data %>% skim() %>% partition()

# descr_stats$logical %>% print_kable("Binary Variables")
descr_stats$factor %>% mutate_if(is.numeric, format_number) %>% print_kable("Factor Variables")
Factor Variables
skim_variable n_missing complete_rate ordered n_unique top_counts
YEARSEQ 0 1 FALSE 118,556 -94: 1, 000: 1, 000: 1, 000: 1
STATE 0 1 FALSE 25 CA: 29564, NY: 17167, FL: 12520, TX: 8965
COUNTY 0 1 FALSE 65 Los: 10419, Coo: 5738, Dad: 4355, Mar: 4245
TOTCHGS1 46 1 FALSE 8 One: 51752, Two: 32156, Thr: 16085, Fou: 8376
OFFENSE1 0 1 FALSE 16 Oth: 20746, Dru: 20534, Ass: 14470, Lar: 11178
OFFTYPE1 0 1 FALSE 4 Dru: 41280, Pro: 37280, Vio: 29623, Pub: 10373
CHG1ATT 1,425 0.988 FALSE 2 No: 113998, Yes: 3133
OFFENSE2 111 0.999 FALSE 19 Not: 51752, Mis: 20818, Oth: 8221, Dru: 5503
OFFTYPE2 111 0.999 FALSE 7 Not: 51752, Mis: 20818, Pro: 14948, Dru: 13724
CHG2ATT 2,457 0.979 FALSE 3 No: 62783, Not: 51752, Yes: 1564
GENDER 358 0.997 FALSE 2 Mal: 98324, Fem: 19874
CJSTATUS 14,503 0.878 FALSE 2 Non: 66335, Act: 37718
CJRELPTR 18,700 0.842 FALSE 2 No: 85820, Yes: 14036
CJRELPRO 15,311 0.871 FALSE 2 No: 86021, Yes: 17224
CJRELPAR 14,503 0.878 FALSE 2 No: 97902, Yes: 6151
CJRELCUS 14,503 0.878 FALSE 2 No: 102465, Yes: 1588
CJRELDIV 22,431 0.811 FALSE 2 No: 95801, Yes: 324
CJRELFUG 22,431 0.811 FALSE 2 No: 93719, Yes: 2406
SERARR 9,987 0.916 FALSE 3 Fel: 64807, No : 30352, Mis: 13410
PRIARR 14,593 0.877 FALSE 10 0: 30352, 10 : 28036, 1: 9864, 2: 8165
PRIFARR 13,454 0.887 FALSE 10 0: 43799, 1: 13068, 10 : 12238, 2: 10071
PRIMARR 13,769 0.884 FALSE 10 0: 42490, 10 : 13712, 1: 13613, 2: 9852
PRIORFTA 10,220 0.914 FALSE 3 No: 37805, Not: 37117, Yes: 33414
SERCONV 7,076 0.94 FALSE 3 No : 46951, Fel: 44821, Mis: 19708
PRICONV 10,678 0.91 FALSE 10 0: 46951, 1: 14940, 2: 10471, 10 : 8962
PRIFCONV 8,455 0.929 FALSE 10 0: 66818, 1: 16698, 2: 9915, 3: 6053
PRIMCONV 9,482 0.92 FALSE 10 0: 59484, 1: 16358, 2: 10169, 3: 6456
PRIVCONV 10,154 0.914 FALSE 3 Not: 58126, No: 38998, Yes: 11278
PRIPRIS 15,673 0.868 FALSE 10 0: 81920, 1: 10278, 2: 5219, 3: 2663
PRIJAIL 14,190 0.88 FALSE 10 0: 63117, 1: 15344, 2: 8738, 3: 5370
RELDET1 3,530 0.97 FALSE 3 Rel: 70586, Det: 42618, Cas: 1822
RELDET2 3,530 0.97 FALSE 8 Non: 34501, Hel: 33983, Fin: 31828, Den: 7011
PTRMIS 297 0.997 FALSE 3 No: 47933, Not: 47673, Yes: 22653
FTA1 1,535 0.987 FALSE 3 No,: 54220, Not: 47218, Yes: 15583
FTA3 1,559 0.987 FALSE 4 Mad: 54219, Not: 47218, FTA: 11232, FTA: 4328
REGION 263 0.998 FALSE 4 Wes: 39890, Nor: 31666, Sou: 29055, Mid: 17682
APPTYP 0 1 FALSE 4 Non: 51307, Par: 44858, Gov: 21233, Leg: 1158
APPJUD 0 1 FALSE 2 Ele: 96165, App: 22391
RELPOS 0 1 FALSE 2 Yes: 114173, No: 4383
NOSCREEN 0 1 FALSE 2 Scr: 79822, No : 38734
PTRPRG89 0 1 FALSE 2 Pro: 79228, No : 39328
PTRPRG01 0 1 FALSE 2 Pro: 101252, No : 17304
PRSCOMP 23,189 0.804 FALSE 2 No: 66727, Yes: 28640, Bla: 0
WHITE 0 1 FALSE 2 No: 91025, Yes: 27531
BLACK 0 1 FALSE 2 No: 74955, Yes: 43601
OTHER 0 1 FALSE 2 No: 116212, Yes: 2344
HISP 0 1 FALSE 2 No: 93259, Yes: 25297
AGED 1,339 0.989 FALSE 8 25-: 20930, 21-: 19892, 30-: 18542, 18-: 18310
JUVSTAT 0 1 FALSE 2 Adu: 114667, Juv: 3889
FINREL 10,548 0.911 FALSE 2 Not: 76180, Mad: 31828, Bla: 0 
descr_stats$numeric %>% mutate_if(is.numeric, format_number) %>% print_kable("Continuous Variables")
Continuous Variables
skim_variable n_missing complete_rate mean sd p0 p25 p50 p75 p100 hist
YEAR 0 1 1,997 4.54 1,990 1,994 1,998 2,002 2,004 ▇▃▇▃▇
AGE 1,341 0.989 30 9.98 12 22 28 36 90 ▇▇▂▁▁
BAILAMT 54,656 0.539 36,094 157,465 1 3,000 10,000 25,000 10,000,000 ▇▁▁▁▁
POPULAT 0 1 2,324,192 2,408,543 585,000 866,000 1,472,000 2,347,000 9,917,300 ▇▁▁▁▁
POPU2 0 1 2,281,627 2,379,665 79,000 858,000 1,446,200 2,301,000 9,763,800 ▇▂▁▁▁
CASES 0 1 1,909 1,277 240 1,082 1,572 2,231 6,400 ▇▆▁▁▁
CASRAT 0 1 0.001 0.001 0 0.001 0.001 0.001 0.007 ▇▁▁▁▁
UCRINDX 0 1 145,509 154,305 2,346 57,126 79,285 184,270 711,295 ▇▂▁▁▁
UCRINDX2 0 1 154,654 164,141 2,815 58,167 83,402 187,048 719,401 ▇▂▁▁▁
CRMRT 0 1 0.067 0.037 0.003 0.043 0.063 0.078 0.309 ▇▆▁▁▁
CRMRT2 0 1 0.074 0.058 0.003 0.047 0.067 0.084 0.766 ▇▁▁▁▁
CRMRTCHG 0 1 -0.007 0.044 -0.696 -0.01 -0.004 0.001 0.149 ▁▁▁▁▇
DELCRMRT 0 1 -3.687 40.4 -90.863 -14.157 -6.487 1.51 788 ▇▁▁▁▁
ADP 1,128 0.99 173 85 2 102 187 254 280 ▃▃▃▅▇
ADP2 1,128 0.99 175 85.2 2 106 191 256 283 ▃▃▃▅▇
RATCAP 1,128 0.99 156 76.8 2 91 172 230 253 ▃▃▃▅▇
RATCAP2 1,128 0.99 159 76.9 2 92 175 231 257 ▃▃▃▅▇
JALAVAL 1,128 0.99 146 85 2 76 143 219 292 ▇▇▇▇▇
JALAVAL2 1,128 0.99 146 83.8 2 78 140 220 290 ▇▇▇▆▇
DELTJLAVL 1,128 0.99 146 81.5 1 74 147 215 292 ▆▇▆▇▆
WHTP 0 1 1,075,965 875,226 107,284 535,609 750,341 1,261,087 3,610,800 ▇▃▁▁▂
WHTP2 13,537 0.886 154 86.2 2 76 155 239 278 ▅▅▅▅▇
BLACKP 0 1 398,736 378,252 6,018 111,696 218,653 577,677 1,423,731 ▇▃▁▂▁
BLACKP2 0 1 393,123 378,608 5,694 105,137 212,971 568,349 1,423,731 ▇▃▁▂▁
ASIANP 0 1 196,116 320,307 3,663 22,984 62,447 210,312 1,321,268 ▇▁▁▁▁
HISPP 0 1 664,536 1,082,081 3,227 63,668 239,483 734,569 4,610,627 ▇▁▁▁▁
HISPP2 13,537 0.886 164 83.4 2 95 175 239 278 ▃▃▅▅▇
PCTBLCK 0 1 0.2 0.146 0.008 0.095 0.169 0.264 0.666 ▇▆▁▃▁
PCTHSP 0 1 0.203 0.158 0.005 0.064 0.174 0.29 0.807 ▇▆▂▁▁
PCTBLCK2 0 1 0.199 0.145 0.008 0.095 0.164 0.263 0.669 ▇▆▁▃▁
PCTHSP2 13,537 0.886 149 77.1 2 82.5 155 221 263 ▅▅▆▆▇
PCTBCHG 0 1 0.001 0.023 -0.355 -0.001 0.001 0.005 0.085 ▁▁▁▁▇
PCTHCHG 13,537 0.886 152 77.6 2 93 158 219 279 ▆▅▇▇▇
DELTBPOP 0 1 1.25 10.3 -91.54 -0.593 1.05 3.01 124 ▁▁▇▁▁
DELTHPOP 13,537 0.886 130 80.2 1 62 122 197 279 ▇▇▇▆▆
EXPND 19,965 0.832 52,260,613 68,656,458 439,535 13,572,816 28,200,000 50,549,963 373,000,000 ▇▁▁▁▁
FULEMP 19,493 0.836 99.9 50.1 2 57 103 146 171 ▃▅▅▆▇
PRTEMP 19,493 0.836 14.4 15.8 2 2 5 26 51 ▇▁▂▁▁
POLCRT 0 1 0.286 0.479 0 0.05 0.18 0.44 8.96 ▇▁▁▁▁
POLJAL 0 1 0.882 1.19 0 0 0.37 1.27 9.96 ▇▁▁▁▁
POLFT_PT 0 1 31.1 12.5 9.23 22.3 28.1 35.9 90.5 ▅▇▁▁▁
POLTOT 31,534 0.734 135 66.4 2 79 145 202 219 ▃▃▃▃▇
POLRTC 31,534 0.734 127 68.5 2 70 126 191 237 ▆▅▇▆▇
POLRT 8,856 0.925 22.3 13.9 2 8 21 33 51 ▇▇▅▅▃
DRGPOL 8,856 0.925 21.8 10.2 2 14 23 30 38 ▆▆▅▇▇
PRCP 19,965 0.832 145 72.4 2 85 154 209 257 ▅▅▆▇▇
PRCPEMP 19,493 0.836 145 71.9 2 90 154 205 259 ▅▆▆▇▇
BLXDLTB 0 1 0.561 5.67 -91.54 0 0 0 124 ▁▁▇▁▁
HPXDLTH 2,584 0.978 27.8 42.1 1 13 13 13 257 ▇▁▁▁▁
BLXPCTB 0 1 0.092 0.15 0 0 0 0.169 0.666 ▇▂▁▁▁
HPXPCTH 0 1 0.066 0.146 0 0 0 0 0.807 ▇▁▁▁▁
PCTBLKSQ 0 1 0.061 0.081 0 0.009 0.029 0.07 0.443 ▇▁▁▁▁
PCTHSPSQ 0 1 0.066 0.092 0 0.004 0.03 0.084 0.651 ▇▂▁▁▁
BAILLN 54,656 0.539 406 222 2 217 429 574 1,000 ▆▇▇▅▁
PBLKLN 0 1 -1.944 0.904 -4.793 -2.349 -1.777 -1.332 -0.407 ▁▂▃▇▃
PHSPLN 0 1 -2.054 1.14 -5.316 -2.747 -1.75 -1.237 -0.214 ▁▂▃▇▅
PRCPLN 19,965 0.832 144 72 2 85 153 208 256 ▅▅▆▇▇
JLAVLLN 1,128 0.99 144 84.1 2 74 140 216 289 ▇▇▇▇▇
CRMRTLN 0 1 -2.834 0.514 -5.928 -3.136 -2.771 -2.554 -1.174 ▁▁▃▇▁
CASRATLN 0 1 -6.932 0.547 -8.625 -7.35 -6.926 -6.535 -4.903 ▁▆▇▂▁
AGESQ 1,339 0.989 20 9.99 2 12 18 26 79 ▇▇▂▁▁

Correlation Matrix

analysis_data %>% mutate_all(as.numeric) %>% cor(use = "pairwise.complete.obs") %>% ggcorrplot(p.mat = cor_pmat(.), hc.order = TRUE)

# p.mat = cor_pmat(.), hc.order = TRUE

Flight Risk Accuracy

Here we are trying to estimate the “accuracy” of those determining whether a person charged with a crime is a flight risk, and therefore should be detained until their trial. While we obviously cannot know whether someone who was detained until trial would have returned for their trial or not from this data, we can attempt to measure this from those who were released after paying their bail. Someone who was released on bail would likely have been considered a significantly greater flight risk than a person released with no bail. We can attempt to measure the amount of false positives by looking at people who were deemed a high enough flight risk for bail, but still returned. Those who were released on bail and did not return are considered to be “correctly” classified here. If someone was released and returned then they were correctly classified, while those who were released without bail and did not return would be considered a false negative. This is not a perfect measurement of classification accuracy, since bail is a variable amount, and is different than being released even if a high flight risk with no collateral, but should be close enough to determine how accuracte and equitable humans are in this process.

tibble(" " = c("Paid Bail", "No Bail"), "Appeared at Court" = c("False Positive", "Correct"), "Did Not Appear" = c("Correct", "False Negative")) %>% print_kable()
Appeared at Court Did Not Appear
Paid Bail False Positive Correct
No Bail Correct False Negative 
accuracy_data = analysis_data %>% 
  filter(RELDET1 == "Released") %>% 
  filter(!is.na(FTA1)) %>%
  mutate(no_bail = (FINREL == "Not made bail"), 
         fta = FTA1 == "Yes, FTA", 
         pred_acc = ifelse(fta, 
                           ifelse(no_bail, "False Negative", "Correct"), 
                           ifelse(no_bail, "Correct", "False Positive")),
         race = ifelse(WHITE == "Yes", "White", ifelse(BLACK == "Yes", "Black", ifelse(HISP == "Yes", "Latinx", "Other"))))

# accuracy_data %>% select(FINREL, no_bail, fta, FTA1, pred_acc) %>% distinct() %>% View()

accuracy_data %>% count(pred_acc) %>% na.omit() %>% mutate(percent = scales::percent(n / sum(n), accuracy = 0.1)) %>% select(-n) %>% rename("Prediction Accuracy" = pred_acc) %>% print_kable("Overall Accuracy")
Overall Accuracy
Prediction Accuracy percent
Correct 47.8%
False Negative 13.4%
False Positive 38.9% 
accuracy_data %>% count(GENDER, pred_acc) %>% na.omit() %>% crosstab_percent(formatted = TRUE) %>% select(-n) %>% pivot_wider(names_from = pred_acc, values_from = percent) %>% rename("Gender" = GENDER) %>% print_kable("Accuracy by Gender")
## Warning: Factor `GENDER` contains implicit NA, consider using
## `forcats::fct_explicit_na`
Accuracy by Gender
Gender Correct False Negative False Positive
Male 46.8% 13.4% 39.8%
Female 51.2% 13.3% 35.5% 
accuracy_data %>% count(AGED, pred_acc) %>% na.omit() %>% crosstab_percent(formatted = TRUE) %>% select(-n) %>% pivot_wider(names_from = pred_acc, values_from = percent) %>% rename("Age Category" = AGED) %>% print_kable("Accuracy by Age Category")
## Warning: Factor `AGED` contains implicit NA, consider using
## `forcats::fct_explicit_na`
Accuracy by Age Category
Age Category Correct False Negative False Positive
1-17 56% 14.4% 29.6%
18-20 48.9% 12.4% 38.8%
21-24 46.1% 12.7% 41.2%
25-29 46.7% 14% 39.3%
30-34 47% 14.4% 38.5%
35-39 47.1% 14.5% 38.4%
40-49 47.7% 13.7% 38.6%
50+ 49.4% 8.6% 42% 
accuracy_data %>% count(race, pred_acc) %>% na.omit() %>% crosstab_percent(formatted = TRUE) %>% select(-n) %>% pivot_wider(names_from = pred_acc, values_from = percent) %>% rename("Race" = race) %>% print_kable("Accuracy by Race")
Accuracy by Race
Race Correct False Negative False Positive
Black 47.4% 14.9% 37.7%
Latinx 48.8% 15.5% 35.7%
Other 48.1% 12.5% 39.4%
White 47.2% 10.5% 42.3% 
# accuracy_data %>% count(c_charge_degree, pred_acc) %>% crosstab_percent(formatted = TRUE) %>% select(-n) %>% pivot_wider(names_from = pred_acc, values_from = percent) %>% rename("Charge Degree" = c_charge_degree) %>% print_kable("COMPAS Accuracy by Charge Degree")
# 
# collapse_by = function(x, collapse_func = quantile, inclusive = TRUE) {
#   raw_values = collapse_func(x)
#   out = character()
#   if (inclusive) {
#     indices = 1:(length(raw_values) - 1)
#     for (val_index in indices) {
#       out[x >= raw_values[val_index] & x <= raw_values[val_index + 1]] = ifelse(val_index == length(indices), paste("<=", (raw_values[val_index + 1])), paste("<", (raw_values[val_index + 1])))
#     }
#   }
#   factor(out)
# }
# 
# accuracy_data %>% mutate(priors_count = collapse_by(priors_count)) %>%  count(priors_count, pred_acc) %>% crosstab_percent(formatted = TRUE) %>% select(-n) %>% pivot_wider(names_from = pred_acc, values_from = percent) %>% rename("Priors Count" = priors_count) %>% print_kable("COMPAS Accuracy by Priors Count")
# 
# accuracy_data %>% filter(!is.na(offense_year)) %>% count(offense_year, pred_acc) %>% crosstab_percent(formatted = TRUE) %>% select(-n) %>% pivot_wider(names_from = pred_acc, values_from = percent) %>% rename("Year" = offense_year) %>% print_kable("COMPAS Accuracy by Year of Offense")
# 
# accuracy_data %>% filter(!is.na(offense_month)) %>% count(offense_month, pred_acc) %>% crosstab_percent(formatted = TRUE) %>% select(-n) %>% pivot_wider(names_from = pred_acc, values_from = percent) %>% rename("Month" = offense_month) %>% print_kable("COMPAS Accuracy by Month of Offense")

Equity Tables

equity_table = function(data_set, demo_var, recid_var, demo_label = str_to_sentence(demo_var), recid_label = "Rate", format_perc = TRUE) {
  data_set %>% 
    count(!!sym(demo_var), !!sym(recid_var)) %>% 
    na.omit() %>% 
    crosstab_percent(vars = demo_var, formatted = format_perc) %>% 
    select(-n) %>% 
    pivot_wider(names_from = recid_var, values_from = "percent") %>% 
    select(-`FALSE`) %>% (function(x) {names(x) = c(demo_label, recid_label); x})
}

equity_data = analysis_data %>% 
  mutate(no_bail = (FINREL == "Not made bail"), 
         fta = FTA1 == "Yes, FTA", 
         race = ifelse(WHITE == "Yes", "White", ifelse(BLACK == "Yes", "Black", ifelse(HISP == "Yes", "Latinx", "Other"))))

Release Rates

equity_data %>% 
  mutate(RELDET1 = RELDET1 == "Released") %>% 
  count(RELDET1) %>% 
  na.omit() %>% 
  mutate(Percent = scales::percent(n / sum(n), accuracy = 0.1), RELDET1 = ifelse(RELDET1, "Yes", "No")) %>%
  rename("Released" = RELDET1) %>% 
  select(-n) %>% 
  print_kable("Overall Release Rate")
Overall Release Rate
Released Percent
No 38.6%
Yes 61.4% 
equity_data %>% 
  mutate(RELDET1 = RELDET1 == "Released") %>% 
  equity_table("GENDER", recid_var = "RELDET1") %>% 
  print_kable("Release Rate by Gender")
## Warning: Factor `GENDER` contains implicit NA, consider using
## `forcats::fct_explicit_na`
## Note: Using an external vector in selections is ambiguous.
## i Use `all_of(recid_var)` instead of `recid_var` to silence this message.
## i See <https://tidyselect.r-lib.org/reference/faq-external-vector.html>.
## This message is displayed once per session.
Release Rate by Gender
Gender Rate
Male 58.9%
Female 73% 
equity_data %>% 
  mutate(RELDET1 = RELDET1 == "Released") %>%  
  equity_table("AGED", recid_var = "RELDET1", demo_label = "Age Group") %>% 
  print_kable("Release Rate by Age Group")
## Warning: Factor `AGED` contains implicit NA, consider using
## `forcats::fct_explicit_na`
Release Rate by Age Group
Age Group Rate
1-17 68.9%
18-20 66.9%
21-24 61.3%
25-29 60.1%
30-34 58.3%
35-39 58.4%
40-49 58.9%
50+ 67.5% 
equity_data %>% 
  mutate(RELDET1 = RELDET1 == "Released") %>% 
  equity_table("race", recid_var = "RELDET1") %>% 
 print_kable("Release Rate by Race")
Release Rate by Race
Race Rate
Black 60.4%
Latinx 53.9%
Other 64.8%
White 67.1%

Bail Rates

bail_data = equity_data %>% filter(RELDET2 == "Financial release" | RELDET2 == "Held on bail" | RELDET2 == "Denied bail") %>% mutate(offered_bail = RELDET2 != "Denied bail")

On average, 90.4% of people charged with a crime and were not released, were offered to be released on bail.

bail_data %>% filter(!is.na(GENDER)) %>% group_by(GENDER) %>% summarise("Rate" = mean(offered_bail) %>% scales::percent(accuracy = 0.1)) %>% rename("Gender" = GENDER) %>% print_kable("Bail Rate by Gender")
Bail Rate by Gender
Gender Rate
Male 89.9%
Female 93.2% 
bail_data %>% filter(!is.na(AGED)) %>% group_by(AGED) %>% summarise("Rate" = mean(offered_bail) %>% scales::percent(accuracy = 0.1)) %>% rename("Age Category" = AGED) %>% print_kable("Bail Rate by Age Category")
Bail Rate by Age Category
Age Category Rate
1-17 87.6%
18-20 91.3%
21-24 90.6%
25-29 89.6%
30-34 89.5%
35-39 90.4%
40-49 90.9%
50+ 92.0% 
bail_data %>% filter(!is.na(race)) %>% group_by(race) %>% summarise("Rate" = mean(offered_bail) %>% scales::percent(accuracy = 0.1)) %>% rename("Race" = race) %>% print_kable("Bail Rate by Race")
Bail Rate by Race
Race Rate
Black 88.9%
Latinx 90.4%
Other 92.5%
White 91.0%

Bail Amount

bail_amt_data = equity_data %>% filter(!(is.na(BAILAMT)))

The median bail amount overall is $10,000.

bail_amt_data %>% select(GENDER, BAILAMT) %>% na.omit() %>% group_by(GENDER) %>% summarise(BAILAMT = median(BAILAMT)) %>% rename("Gender" = GENDER, "Median Bail" = BAILAMT) %>% mutate_if(is.numeric, format_number) %>% print_kable("Median Bail Amount by Gender")
Median Bail Amount by Gender
Gender Median Bail
Male 10,000
Female 5,000 
bail_amt_data %>% select(AGED, BAILAMT) %>% na.omit() %>% group_by(AGED) %>% summarise(BAILAMT = median(BAILAMT)) %>% rename("Age Category" = AGED, "Median Bail" = BAILAMT) %>% mutate_if(is.numeric, format_number) %>% print_kable("Median Bail Amount by Age Category")
Median Bail Amount by Age Category
Age Category Median Bail
1-17 5,000
18-20 7,500
21-24 10,000
25-29 10,000
30-34 10,000
35-39 10,000
40-49 10,000
50+ 10,000 
bail_amt_data %>% select(race, BAILAMT) %>% na.omit() %>% group_by(race) %>% summarise(BAILAMT = median(BAILAMT)) %>% rename("Race" = race, "Median Bail" = BAILAMT) %>% mutate_if(is.numeric, format_number) %>% print_kable("Median Bail Amount by Race")
Median Bail Amount by Race
Race Median Bail
Black 7,500
Latinx 10,000
Other 8,000
White 7,500